# A Feast for Throws/December 5, 2018by Eliot McKinley

By Eliot McKinley (@etmckinley)

In Game of Throw-Ins, I characterized and introduced an expected throw-in possession retention model (xRetain) for MLS. Go read the whole thing, but it showed that throw-ins are more likely to be completed and possession retained when they are thrown backwards, quickly, and outside a team’s defensive third. But what are MLS teams and players doing with their throw-ins?

To help differentiate teams’ throw-in styles, I turned to hierarchical clustering (see the graph below). I won’t get into mathematical details, but you can think of it sort of like an evolutionary tree. However, instead of the branches separating species, they are separating different throw-in angle frequencies. Kind of like how humans and chimpanzees are near each other on the branches of an evolutionary tree but far away from birds, teams which always throw the ball backwards and short will be far away from those that always take throw-ins forward and long.

Looking at just the 2018 season, I cut the tree (red line) into four general groups and showed each group’s Thrownar (it’s like PassSonar, but for throw-ins, get it?). Each bar on the Thrownar represents the relative frequency of throw-ins taken at each 15-degree interval and the color is the average distance for each. So the most frequent angle will always be the longest bar and the rest are scaled to that.

Overall, teams break down into those that mostly take vertical throws-ins and those that will mix it up with some backwards and horizontal throws as well. Most teams are in the former category, but six courageous teams do something a bit different. In group one, Philadelphia, Columbus, and Orlando most commonly throw backwards from the left side of the field, and rarely throw vertically from the right. Group two (New York City, LAFC, and Kansas City) are similar to group one, as you’d expect from the tree, but are more balanced with their long vertical throws on either side and also have higher frequencies of lateral throw-ins. Again they throw backwards more from the left than from the right. I can’t think of a good reason why throw-ins are more likely to go backwards on the left side of the field than the right. Maybe it has something to do with most players being right footed? If you have any ideas please let me know.

On the other side of the tree, groups three and four represent the teams that predominantly throw vertically. Both show similar patterns, except group four takes 33% of their throws almost directly forward, whereas for group three it is 23%. Group four includes New York Red Bulls and New England, who apparently translate their typical direct passing style to throw-ins by very rarely going backwards.

We can go even deeper and look and look at throw-ins by which third of the field they are taken from. Generally, most teams are very similar with their throw-ins in the middle third of the field and in the defensive third all teams (except one) have similar profiles. But in the attacking third there is a difference between teams that take long throws into the box and those that don’t, Sporting Kansas City in 2018 did it’s own thing in their defensive third. I already tweeted the tree, but here are the Thrownars. Of the 85 team seasons since 2015, 84 have taken throw-ins in the defensive third by launching the ball downfield, but SKC did it very differently this year, generating a much more even distribution of throw-in angle frequencies. And what Kansas City did this season worked. In previous seasons, Kansas City was a middling to poor team at retaining possession of throw-ins in their defensive third, having a throw-in retention Per100 of -2.9 (Per100 is the number of throw-ins retained compared to the expected number of throw-ins retained, normalized to 100 passes). However, in 2018, SKC’s Per100 in the defensive third was 7.1. Not bad, especially for a team that prioritizes possession like Peter Vermes’ SKC did this season.

SKC showed that by changing strategies a team can improve throw-in retention, but what we really want to know is which teams are the best at keeping possession of their throw-ins. It turns out that, at least in regards to throw-ins, the United States Soccer Federation made an inspired choice in Gregg Berhalter as manager of the men’s national team. Berhalter’s Columbus Crew have consistently been one of the best teams in the league at keeping the ball after throw-ins and have improved every season since 2015. Peter Vermes has seen his team improve season to season, while Brian Schmetzer’s Seattle has gotten worse over time. San Jose fans can take heart that while Mikael Stahre may have been a failure overall, at least he improved the team’s throw-in retention.

Finally, if you have a throw in and you absolutely need to retain possession, who do you call on? Harrison Afful of course. Over the last two seasons throw-ins from Afful have been retained by the Crew 38 more times than expected. On the opposite side is DC’s Nick Deleon, who retained the ball 26 times fewer than expected. Much of this may be down to systems, as Chris Klute, Corey Ashe, Waylon Francis, and Hernan Grana showed increased Per100 with Columbus compared with their other teams.

In summary, while there are many similarities amongst MLS teams, there are teams and players that are able to differentiate themselves in both style and execution. Sporting Kansas City totally revamped how they took throw-ins in their defensive third and saw improvements in possession retention. Gregg Berhalter should seemingly be holding coaching clinics on throw-ins along with his long time right back Harrison Afful. Of course, retaining possession after a throw-in may not be what a team is intending to do, necessarily. The ultimate end product in soccer is a goal, and for some teams, such as RBNY, possession isn’t what gets them goals and their throw-in style seems to support that. In the final installment of my throw-in trilogy, I’ll look into how teams convert their throw-ins to shots and goals.